摘要
提出了1种基于综合特征的花卉识别方法﹒该方法首先利用图像显著性进行分割,以实现前景背景分离;然后分别提取花卉的颜色特征、形状特征和纹理特征,在提取纹理特征时,为了提高特征对花卉的表述能力,对图像进行边缘增强和压缩处理;最后使用SVM分类器进行分类识别﹒实验分别与BP神经网络、KNN最近邻分类这2种分类方法进行了对比分析,相对于BP神经网络的分类识别率(85.81%)和KNN最近邻分类的识别率(84.09%),基于综合特征的识别方法具有更高的准确率,识别率可以达到93.7%﹒
A method of flower species identification based on integrated features is proposed.Firstly,the image saliency is used to segment foreground and background.Then the color features,shape features and texture features of flowers are extracted respectively.In order to improve the expressive ability of the characters to the flowers,the texture features are extracted and compressed.At last,the SVM classifier is used for classification recognition.The experiment is compared with the 2 classification methods of BP neural network and KNN nearest neighbor classification.Compared with the classification recognition rate(85.81%)of BP neural network and the recognition rate of the KNN nearest the neighbor classification(84.09%),the recognition method based on the comprehensive feature has a higher accuracy rate and the recognition rate can reach 93.7%.
作者
王威
刘小翠
王新
WANG Wei;LIU Xiaocui;WANG Xin(School of Computer and Communication Engineering,Changsha University of Science&Technology,Changsha,Hunan 410114,China)
出处
《湖南城市学院学报（自然科学版）》
CAS
2018年第4期45-49,共5页
Journal of Hunan City University:Natural Science
基金
湖南省教育厅科研项目(17C0043)
湖南省军民融合协同创新项目(2017)
